Direct dating of remains from the site of Vindija and implications for the Middle to Upper transition

Thibaut Devièsea,1, Ivor Karavanicb,c, Daniel Comeskeya, Cara Kubiaka, Petra Korlevicd, Mateja Hajdinjakd, Siniša Radovice, Noemi Procopiof, Michael Buckleyf, Svante Pääbod, and Tom Highama

aOxford Radiocarbon Accelerator Unit, Research Laboratory for and the History of Art, University of Oxford, Oxford OX1 3QY, ; bDepartment of Archaeology, Faculty of Humanities and Social Sciences, University of Zagreb, HR-10000 Zagreb, ; cDepartment of Anthropology, University of Wyoming, Laramie, WY 82071; dDepartment of Evolutionary Genetics, Max-Planck-Institute for Evolutionary Anthropology, D-04103 Leipzig, ; eInstitute for Quaternary Palaeontology and Geology, Croatian Academy of Sciences and Arts, HR-10000 Zagreb, Croatia; and fManchester Institute of Biotechnology, University of Manchester, Manchester M1 7DN, United Kingdom

Edited by Richard G. Klein, Stanford University, Stanford, CA, and approved July 28, 2017 (received for review June 5, 2017) Previous dating of the Vi-207 and Vi-208 Neanderthal remains from to directly dating the remains of late and early (Croatia) led to the suggestion that Neanderthals modern , as well as artifacts recovered from the sites they survived there as recently as 28,000–29,000 B.P. Subsequent dating occupied. It has become clear that there have been major pro- yielded older dates, interpreted as ages of at least ∼32,500 B.P. We blems with dating reliability and accuracy across the Paleolithic have redated these same specimens using an approach based on the in general, with studies highlighting issues with underestimation extraction of the amino acid hydroxyproline, using preparative high- of the ages of different dated samples from previously analyzed performance liquid chromatography (Prep-HPLC). This method is sites (6). We have been working on redating some of the pur- more efficient in eliminating modern contamination in the bone col- ported late-surviving Neanderthal sites from around Europe, lagen. The revised dates are older than 40,000 B.P., suggesting the which have included and archaeological remains from Vindija Neanderthals did not live more recently than others across sites such as Mezmaiskaya (), where a previous directly

Europe, and probably predate the arrival of anatomically modern dated Neanderthal infant yielded a radiocarbon age of ∼29,000 ANTHROPOLOGY humans in Eastern Europe. We applied zooarchaeology by mass B.P. (7), and Zafarraya (), which was thought to contain spectrometry (ZooMS) to find additional hominin remains. We iden- Neanderthal remains clustering in age around a small group of tified one bone that is Neanderthal, based on its mitochondrial DNA, U-series–dated animal bones between 33,400 and 28,900 B.P. and dated it directly to 46,200 ± 1,500 B.P. We also attempted to (8). At Mezmaiskaya, the AMS dates obtained for the Nean- date six early Upper Paleolithic bone points from stratigraphic units derthal excavated above the previously dated individual were + ± G1,Fd/d G1 and Fd/d, Fd. One bone artifact gave a date of 29,500 substantially older (9). This, along with other AMS dates from 400 B.P., while the remainder yielded no collagen. We additionally cut-marked fauna from the same archaeological horizons, sug- GENETICS – dated animal bone samples from units G1 and G1 G3. These dates gested the original date of 29,000 B.P. could not be correct. At suggest a co-occurrence of early Upper Paleolithic osseous artifacts, Zafarraya, Wood et al. (10) showed that, when redated using particularly split-based points, alongside the remains of Neander- ultrafiltration methods, the bones that produced ages of ∼33,000 thals is a result of postdepositional mixing, rather than an associa- B.P. were in fact beyond the radiocarbon limit, suggesting the tion between the two groups, although more work is required to show this definitively. Significance

Vindija Cave (Croatia) | single-compound AMS dating | Radiocarbon dating of Neanderthal remains recovered from DNA analysis | zooarchaeology by mass spectrometry | Vindija Cave (Croatia) initially revealed surprisingly recent re- Middle to Upper Paleolithic transition sults: 28,000–29,000 B.P. This implied the remains could repre- sent a late-surviving, refugial Neanderthal population and ∼ he period between 45,000 and 35,000 cal B.P. in Europe suggested they could have been responsible for producing Twitnessed the so-called biocultural transition from the Mid- some of the early Upper Paleolithic artefacts more usually produced dle to early Upper Paleolithic, when incoming anatomically by anatomically modern humans. This article presents revised ra- modern humans displaced Neanderthal groups across the con- diocarbon dates of the human bones from this site obtained using tinent (1, 2). Significant questions still remain regarding the a more robust purification method targeting the amino acid precise nature of this transition, the humans responsible for the hydroxyproline. The data show that all the Neanderthal remains various transitional early Upper Paleolithic industries, the de- are from a much earlier period (>40,000 cal B.P.). These revised gree of overlap between Neanderthals and modern humans, and dates change our interpretation of this important site and dem- the timing of the disappearance of the former. The European onstrate that the Vindija Neanderthals probably did not overlap record for the transition retains its interest because it is the best- temporally with early modern humans. documented sequence for the disappearance of a hominin group available (3). The latest data, both radiometric and genetic, Author contributions: T.D., S.P., and T.H. designed research; T.D., I.K., D.C., C.K., P.K., M.H., S.R., N.P., M.B., S.P., and T.H. performed research; T.D., D.C., and T.H. contributed suggest Neanderthals and modern humans coexisted or over- new reagents/analytic tools; T.D., I.K., D.C., C.K., P.K., M.H., S.R., N.P., M.B., S.P., and T.H. lapped for up to several thousand years in Europe until Neander- analyzed data; T.D. and T.H. wrote the paper; and I.K., D.C., C.K., P.K., M.H., S.R., N.P., thal disappearance at around 40,000 cal B.P. (4, 5). Ascertaining the M.B., and S.P. contributed to writing the paper. spatial attributes of Neanderthal and modern human populations The authors declare no conflict of interest. in Europe is an area of active research, and a reliable chronology This article is a PNAS Direct Submission. remains essential. Freely available online through the PNAS open access option. Our understanding of the biocultural processes involved in the 1To whom correspondence should be addressed. Email: [email protected]. transition have been greatly influenced by improved accelerator This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. mass spectrometry (AMS) dating methods and their application 1073/pnas.1709235114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1709235114 PNAS Early Edition | 1of6 Downloaded by guest on September 27, 2021 Neanderthal remains were unlikely to be as young as previously Svoboda (18)]. Karavanic and Smith (19) have suggested that the thought. In both cases, revised radiocarbon dates produced with mixture of elements may represent the interaction and possible more robust chemical pretreatment methods have illustrated sig- acculturation between modern humans and late Neanderthals. nificant underestimates in the previous dates that cannot be rec- These alternatives are testable by selecting human and organic onciled with a hypothesis of late-surviving refugial Neanderthals. osseous points, as well as animal bones, for renewed AMS dating. The Neanderthal fossil remains from level G1 of Vindija Cave This is what we have undertaken and describe here. in northern Croatia have remained in the literature as potentially late individuals. Given the evidence from the Pes¸tera cu Oase Materials specimen, which demonstrates a recent Neanderthal ancestry in A total of 10 samples from Vindija Cave were selected for AMS radiocarbon a 40,000 cal B.P. modern human from the Danube corridor (5), dating (Table 1). These included three previously dated Neanderthal specimens the renewed dating of the Vindija remains is overdue. (Vi-207, Vi-208, and Vi-33.19), as well as a fourth Neanderthal bone (Vi-*28) Two specimens, Vi-207 and Vi-208, were originally directly discovered using zooarchaeology by mass spectrometry (ZooMS) screening. In AMS dated in the late 1990s at the Oxford Radiocarbon Ac- addition, to test the reality of the co-occurrence of earlier Upper Paleolithic bone and antler point artifacts with the Neanderthal remains, we selected six celerator Unit (ORAU). Vi-207 is a right posterior mandible and osseous points for dating (SI Appendix,Fig.S1). To test collagen preservation, Vi-208 is a parietal fragment, both showing Neanderthal-specific we took ∼3–5 mg bone powder, using tungsten carbide drills, and measured morphology (11, 12). The initial radiocarbon results were the %N content. This is an indicator of collagen preservation (20). The results 29,080 ± 400 B.P. (OxA-8296) and 28,020 ± 360 B.P. (OxA- show that only one bone point sample (Vi-3446) had sufficient levels of ni- 8295) (13). Higham et al. (14) attempted to redate these speci- trogen to warrant full sampling for collagen extraction and AMS dating; the mens by taking the very small amounts of collagen remaining remainder failed and therefore were not sampled further (SI Appendix,Table from the original sample pretreatment and ultrafiltering the S1). We also included the sample from a split-based bone point (Vi-3437) that product before AMS dating. The revised measurements were had been analyzed in the laboratory previously, producing only a small collagen 32,400 ± 1,800 B.P. (Vi-207: OxA-X-2089-07) and 32,400 ± 800 yield. We decided to attempt to redate it, using a larger starting mass of bone B.P. (Vi-208: OxA-X-2089-06), which indicated the previous powder. Unfortunately, there was insufficient collagen remaining from this sample after pretreatment. We also selected four animal bones of Cervidae dates were indeed too young. For sample Vi-208, after ultrafil- (Vi-*17), Panthera sp. (Vi-*6), Ursus sp. (Vi-*7), and Bovidae (Vi-*60), iden- tration, the C/N atomic ratio was 3.4, which indicates collagen of tified by ZooMS from stratigraphic units G1 and G1–G3 to explore the po- acceptable quality. However, for Vi-207, the >30-kDa fraction tential issue of postdepositional mixing in the deposits further. obtained produced a C/N ratio of 4.3, which indicates the pres- ence of a high molecular weight contaminant. The radiocarbon Results and Discussion date for this sample could therefore include a higher molecular ZooMS Collagen Fingerprinting. We used ZooMS to identify po- weight noncollagenous contaminant, possibly cross-linked to the tential hominin bone fragments among the unidentified faunal collagen. On the basis of the potential problems associated with remains from the G1 and G3 levels, as well as the stratigraphic unit the small size of the redated samples and the potential for re- G1–G3. The majority of the 383 samples we analyzed yielded poor maining contaminants, OxA-X-2089-06 was considered to be a collagen preservation, which prevented any identification to genus minimum age (14). If the dates are even approximately correct, or taxon. Only 101 samples produced identifiable spectra; a sum- however, it makes them the most recent known Neanderthals. mary of all taxa identified by ZooMS is shown in SI Appendix, This would imply a more extensive temporal overlap between Table S2. This assemblage is dominated by Ursus, and only six of Neanderthals and early modern humans in central Europe than the 27 taxa identified by morphological study of the bones in has recently been documented (4). Miracle et al. (21) could be identified here. We identified a single In addition to the Neanderthal remains, level G1 has yielded a hominin specimen (Fig. 1 A and B), which again highlights the use small archaeological assemblage that contains techno-typologically of applying such techniques to groups of unidentified Paleolithic Middle and Upper Paleolithic lithic artifacts plus several distinc- bone samples. The bone was analyzed using ancient DNA tech- tively early Upper Paleolithic osseous points (12). It has been ar- niques to enable a formal species identification. gued that the mix of Neanderthals, Middle Paleolithic tools, and Upper Paleolithic technology was the result of cryoturbation and DNA Analysis of the Human Bones. Genomic analysis based on mi- Ursus spelaeus activity in level G1, with elements mixing into level tochondrial DNA revealed that all four human specimens fall into G1 from both the Upper Paleolithic unit F above and the Middle Neanderthal mitochondrial variation. Full mitochondrial genomes of Paleolithic level G3 below (15, 16). Zilhão (17) has suggested that Vi-207 and Vi-*28 were reconstructed with an average coverage of the G1 lithic assemblage has parallels with the Szeletian tech- 103-fold and 257-fold, respectively. The mitochondrial DNA se- nocomplex, and further, that there is a mixture of elements quence of Vi-207 was identical to Vi-33.25 and Feldhofer 1 mito- of Szeletian and I and II within the level [see also chondrial genomes, whereas Vi-*28 had an identical mitochondrial

Table 1. Samples selected for AMS dating ORAU P Number Sample Reference Archeology Level Species

39039 Vi-33.19 (SP2756) G3 Neanderthal

41681 Vi-*28 (SP4162) G1 Neanderthal

41415 Vi-208/Vi-11.29 (SP3563) G1 Neanderthal 41416 Vi-207/Vi-11.41 (SP3562) G1 Neanderthal 41417 Vi-3446 Fd/d+G1 Not known (point)

41687 Vi-3437 G1 Not known (split-based point)

41803 Vi-*17 G1 Cervidae 41801 Vi-*6 G1 Panthera sp

41802 Vi-*7 G1 Ursus sp

41804 Vi-*60 G1–G3 Bovidae

All samples produced enough collagen for radiocarbon dating with the exception of P41687, which failed to produce any collagen. See SI Appendix, Fig. S1 for images of the bone points included in this study.

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1709235114 Devièse et al. Downloaded by guest on September 27, 2021 Fig. 1. (A) High-resolution photographs of the Vi-*28 Neanderthal bone found using ZooMS. The bone yields evidence for a probable cut and gauge marks (right upper part of the bone). The picture was taken after the bone had undergone sampling for ZooMS and before sampling for aDNA, radiocarbon, and stable isotope analysis. (B) MALDI-TOF mass spectrum of digested collagen from the Vi-*28 bone. All tagged peaks (A, B, C, D, and F) denote sequence- matched peptides observed in human collagen (27, 28).

sequence to Vi-33.17 (SI Appendix,Fig.S2). Both Vi-33.25 and Vi- on the four Neanderthal specimens at the ORAU are reported in 33.17 were found in layer I of Vindija Cave. As previously Table 2. We also list the dates obtained on two other hominin published, Vi-33.19 has the same mitochondrial sequence as Vi- samples: Vi-75-G3/h-203, analyzed at the Uppsala Radiocarbon 33.16 (22). Because of lower endogenous DNA content in Vi- Laboratory (Sweden) (23), and Vi-2291-18 (level G, sublayer un- 208, a full mitochondrial genome could not be reconstructed for known), prepared at the Max Planck Institute, Leipzig, and dated the sample. However, from the limited amounts of mitochondrial at the ORAU (24). The different sample pretreatments are also sequences, we were able to trace most of the observed variants indicated in Table 2. ANTHROPOLOGY to variations found in previously sequenced Neanderthal mito- Vi-208 and Vi-207 produced hydroxyproline dates of 42,700 ± chondrial genomes (SI Appendix, Fig. S3). 1,600 and 43,900 ± 2,000 B.P., respectively. These ages are sig- nificantly older than any of the dates obtained previously for these AMS Dating. Nine of the samples selected produced enough col- specimens using the AG (gelatinized filtered collagen) and AF lagen (or hydroxyproline) to be dated by AMS. All dates obtained (ultrafiltered collagen) procedures, and this strongly suggests that GENETICS

Table 2. Radiocarbon dates of the Vindija Neanderthal remains P no. OxA/OxA-X CRA ± P code Used, mg Yield, mg %Yield %C δ13C, ‰ δ15N, ‰ C/N

Sample Vi-208 (SP3563) 9663 8295 28,020 360 AG 233.9 15.2 6.5 37.1 −19.5 10.6 3.2 9663 2082-09 29,200 360 AG 229.9 10.9 4.7 42.7 −19.8 11.4 3.6 9663 2089-06 32,400 800 AF n/a n/a n/a 42.3 −20.2 10.3 3.4 41415 X-2689-09 42,700 1,600 HYP 626.0 35.1 5.6 45.2 −26.1 10.2 5.6 Sample Vi-207 (SP3562) 9665 8296 29,080 400 AG 229.2 9.7 1.5 36.6 −20.5 11.3 3.6 9665 2082-10 29,100 360 AG 128.8 6.7 5.2 41.6 −22.8 12.2 4.1 9665 2089-07 32,400 1,800 AF n/a n/a n/a 39.0 −24.6 11.1 4.3 41416 X-2689-10 43,900 2,000 HYP 629.0 37.4 6.0 42.2 −24.7 11.5 5.6 Sample Vi-33.19 (SP2756) 39039 32278 45,300 2,300 AF 560 62.2 11.1 45.2 −18.6 10.7 3.4 39039 X-2717-11 44,300 1,200 HYP n/a n/a n/a 46.5 −24.4 12.3 4.9 Sample Vi-*28 (SP4162) 41681 X-2687-57 46,200 1,500 HYP 1,000.0 54.0 5.4 34.3 −25.5 11.7 5.0 Sample Vi-2291–18 (level G, sublayer unknown) 22966 V-2291-18 44,450 550 XB / / / 39.3 −18.7 10.8 3.3 Sample Vi-75-G3/h-203 / Ua-13873 >42,000 −19.4 15.2 3.3

Radiocarbon dates of the Vindija Neanderthal remains. CRA is conventional radiocarbon age, expressed in years B.P. (49). Stable isotope ratios are expressed in per mil (‰) relative to Vienna Pee Dee Belemnite with a mass spectrometric precision of ±0.2‰ (50). PCode refers to pretreatment code; AG is gelatinized filtered collagen, AF is ultrafiltered collagen, XB denotes collagen extracted at a different laboratory than the ORAU, and HYP denotes the extraction of hydroxyproline from hydrolyzed bone collagen (51, 52). Collagen yield represents the weight of gelatin or ultrafiltered gelatin in milligrams. %Yld is the percent yield of extracted collagen as a function of the starting weight of the bone analyzed. When further purification (AF or HYP) was performed on previously extracted collagen, the yields are not reported. %C is the carbon present in the combusted sample (gelatin or hydroxyproline). C/N is the atomic ratio of carbon to nitrogen and is acceptable if it ranges between 2.9–3.5 in the case of collagen or ∼5.0 in the case of hydroxyproline. All samples were prepared and dated at the ORAU with the exception of Vi-2291-18, which was extracted at the Max Planck Institute in Leipzig and dated in Oxford (24) and Vi-75-G3/h-203, which was prepared and dated at Uppsala in Sweden (23). The background carbon derived from the HPLC separation has been accounted for by using a correction where appropriate (details can be found in the SI Appendix). Slash indicates data not available.

Devièse et al. PNAS Early Edition | 3of6 Downloaded by guest on September 27, 2021 Fig. 2. Bayesian age model showing the calibrated HYP ages of the four Neanderthal samples from Vindija Cave. The model is a simple phase model in OxCal 4.3 (47), in which all F14C determinations are assumed to have no relative order. “Start” and “End” correspond to the boundaries calculated by the model. The calibration curve of Reimer et al. (48) was used to calibrate the results. Details can be found in the SI Appendix.

noncollagenous high molecular weight contaminants, probably ultrafiltration (AF). These samples produced dates that were all crosslinked to the collagen, were still present in the sample pre- greater than or very close to the radiocarbon limit (Table 3). These viously dated. It is only by hydrolyzing the collagen and selecting samples are less likely to have been treated with conservation the hydroxyproline that we were able to successfully remove these materials than the human bones, and therefore were not prepared contaminants. using the HPLC method. Three of these faunal bones are from The AMS measurement of the third human bone from level level G1 (Table 1). We see more variation that would be expected G1 (Vi-*28), identified using the ZooMS method, gave a date statistically if we combine the three Neanderthal bones and three ± 2 of 46,200 1,500 B.P. For these three HYP (extraction of faunal bones from level G1 for a χ test (error weighted mean, hydroxyproline from hydrolyzed bone collagen) dates obtained 0.0027 ± 0.0004; t = 22.52; χ2 = 11.07). This shows there is bone of χ2 on the Neanderthal bones from level G1, we performed a test variable radiocarbon age in the G1 level and suggests the possibility using the modern fraction F14C and its error. The error of postdepositional mixing and movement of material. In conse- weighted mean is 0.0038 ± 0.0005 with a t value of 2.57. If t quence, the bone tools cannot be associated with the Neanderthal is <5.99, the value for χ2, the error weighted mean is not sig- remains unless further direct radiocarbon dating using the HYP nificant. This shows that the results obtained on the three approach is undertaken. Clearly, the low collagen content of the Neanderthal bones from level G1 are statistically in agreement. points appears to preclude this at this time. Vi-33.19 (level G3) was also dated using the hydroxyproline method and produced a date of 44,300 ± 1,200 B.P., which is a Conclusions more precise date than the one obtained using the AF procedure Single-amino acid AMS dating of the Vindija Neanderthals has (45,300 ± 2,300 B.P.). The AF determination had a low target yielded results that are substantially older than the previous ages current in the AMS (85% of normal current) and was imprecise that were initially obtained. We have shown that the Neander- for this reason. The fact that the two dates overlap suggests no thals predate ∼44,000 cal B.P. The results suggest this group was significant contamination in this bone. Taken together, these not a late-surviving refugial Neanderthal population, as previously dates show a significantly older occupation of the site by Neander- thought, and means the group almost certainly did not overlap with thals, suggesting the site cannot be considered to be a refugium for early anatomically modern humans in this part of Europe. Despite late-surviving Neanderthals (Fig. 2, code in SI Appendix). The two our best attempts, we were not able to date the bone industry other dates obtained on samples Vi-33.26 and Vi-75, even if ob- associated with the archaeology of level G1. The one date we tained with a method that can be less efficient in removing con- obtained from a later stratigraphic unit was younger than 30,000 tamination, confirm the dating of level G is more than 42,000 B.P. B.P., but because the bone was not treated with the most rig- Only one bone point (Vi-3446–P41417) produced enough ni- orous pretreatment chemistry methods, it could potentially be trogen during prescreening to warrant pretreatment chemistry. older. The dating of other faunal materials from level G1 highlighted Less than 500 mg bone was available, so we applied an AG a significant range in age, which could indicate a perturbation of (gelatin) treatment to maximize the collagen yield. We obtained the general sequence. The question, then, of whether some of a date of 29,500 ± 400 B.P., which we consider to be a minimum the points could have been produced by Neanderthals remains age, as this pretreatment has been shown before to underesti- open; however, it is parsimonious to conclude that the split-based mate the age of ancient bones such as this (Table 3). To test point at least must have a maximum age of 32,000–34,000 B.P. based possible perturbation in level G1, we also dated four faunal on evidence for its association with the Aurignacian in other regions, samples identified as originating from levels G1 and G1–G3,using and so it likely postdates the Vindija Neanderthals significantly.

Table 3. Radiocarbon determinations and analytical data from Vindija faunal remains and bone point OxA/OxA-X CRA ± PCode Species Used, mg Yield, mg %Yield %C δ13C, ‰ δ15N, ‰ C/N

34458 29,500 400 AG Not known (bone point) 450 6.58 1.5 46.0 −19.4 3.5 3.4 2695-21 >46,700 AF Cervidae 960 2.60 0.3 35.9 −20.7 5.6 3.3 34471 >48,400 AF Panthera sp. 1,100 9.09 0.8 39.1 −18.9 10.1 3.2 34472 >49,000 AF Ursus sp. 1,000 4.53 0.5 37.1 −23.0 8.8 3.2 34473 47,200 2,900 AF Bovidae 950 40.41 4.3 42.2 −20.8 4.4 3.2

See SI Appendix, Table S2 caption for details of the terms used.

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1709235114 Devièse et al. Downloaded by guest on September 27, 2021 Bone points have been recovered throughout Eurasia with dates as prepared specifically for this project, whereas Vi-33.19 was previously ana- early as ∼37,000 B.P. (e.g., from the Aurignacian site of Pes’ko and lyzed as part of a separate study (22). Bones were sampled using a sterile the Châtelperronian site of Arcy-sur-Cure) (6, 25, 26). It is there- dentistry drill, and between 15 and 30 mg bone powder was used for DNA fore clear that both anatomically modern humans and Neander- extraction, with the exception of Vi-*28, in which three separate parts of the thals produced bone points, with only split-based bone/antler points bone were sampled (SI Appendix, Table S3). DNA extraction was performed being diagnostic of the earlier facies of the former. using a method for ultrashort DNA fragment retrieval (33) with modifica- tions (34). The bone powder of Vi-207 and Vi-208 was treated with 0.5% Our perception of the biological transition between Nean- hypochlorite solution before DNA extraction (34) to remove microbial and derthals and modern humans has changed radically during the modern human DNA contamination. Next-generation sequencing libraries last decade. Evidence suggests interbreeding and a significant were prepared from 10 μL of each extract, using two single-stranded library ∼ – temporal overlap between the two from 44,000 40,000 cal preparation schemes: Vi-207 and Vi-208 with a previously published method B.P. On the basis of our hydroxyproline dates and the DNA (35), with modifications from Korlevic et al. (34), and Vi-*28 with an auto- results, the Vindija Neanderthals date before the period when mated library preparation scheme (36). Libraries were amplified and labeled the first modern humans arrived into Europe and interbred with a pair of unique index sequences (37, 38). with Neanderthals. Enrichment of the mitochondrial DNA and sequencing. Amplified libraries were enriched for human mitochondrial DNA, using a bead-based hybridization Methods capture (39) with modifications (40). Enriched libraries were sequenced on an Bone Screening by ZooMS. We used collagen peptide mass fingerprinting or Illumina MiSeq platform with a double index configuration (41). Overlapping ZooMS to analyze preserved type 1 collagen (COL1) from small unidentified paired-end reads were merged (42) and mapped to the revised Cambridge bone fragments from the Vindija site. COL1 is characterized by three poly- Reference Sequence mitochondrial genome (rCRS; NC_0120920), using Burrows- peptide (alpha) chains, and within these chains there are small differences in Wheeler Aligner (43) with parameters “-n 0.01 -o 2 -l 16500” (44). Analyses the amino acid sequences among different species. These offer the possibility were restricted to sequences with perfect index combinations. PCR duplicates of identification of species-specific amino acid markers. We used matrix- were removed using bam-rmdup (https://bitbucket.org/ustenzel/biohazard- assisted laser desorption/ionization mass spectrometry to determine the tools), and mapped sequences longer than 35 bases with a mapping quality of relative mass-to-charge (m/z) values of the tryptic digest peptides in each at least 25 were retained for subsequent analyses. To assess whether the results bone sample. By comparing these values with a library of known fauna, we could be affected by present-day modern human contamination, sequences identified bones to genus or species in some instance (27–31). We screened a with ancient DNA specific cytosine to thymine substitutions (C to T) at the 5′ or total of 383 unidentified bone samples from Vindija Cave (levels G1, G3, and 3′ molecule end were selected and analyzed separately. stratigraphic unit G1–G3). Phylogenetic analyses. To reconstruct the mitochondrial genomes of the three previously unanalyzed hominin samples, DNA sequences were realigned to the ANTHROPOLOGY Radiocarbon Dating. We used two different methods to prepare the samples Vindija 33.16 mitochondrial genome (AM948965) (45). Data were processed as for AMS dating. First, samples from artifacts were pretreated following described earlier. Data analysis was performed as previously published (ref. 28 the routine ORAU procedure, comprising a decalcification, base wash, for full mitochondrial genomes, ref. 46 for partial mitochondrial data). reacidification, gelatinization, and ultrafiltration (coded AF in the ORAU), as described by Brock et al. (20). Sample P41417 produced too little collagen for ACKNOWLEDGMENTS. We thank Fred H. Smith (Illinois State University), ultrafiltration, so the treatment was halted after gelatinization (coded AG). Jadranka Mauch Lenardic (Institute for Quaternary Palaeontology and The four human bone samples were dated using the single amino acid ra- Geology, Croatian Academy of Sciences and Arts), and Pavao Rudan (Croatian

diocarbon dating method developed at ORAU (32). This method involves Academy of Sciences and Arts) for their assistance in sampling and analyzing GENETICS separation of the underivatized amino acids from hydrolyzed bone collagen the Neanderthal and faunal bone from Vindija Cave. Ian Cartwright (Institute samples using preparative HPLC. Details of the method are described in the of Archaeology, University of Oxford) took the photographs of the Vi-*28 SI Appendix. Hydroxyproline (HYP), was isolated by Prep-HPLC, combusted, Neanderthal bone. Rachel Hopkins ran the %N tests on some of the bones. graphitized, and AMS-dated. This pretreatment approach is the most effi- We also thank the staff of the Oxford Radiocarbon Accelerator Unit at the cient technique to remove contaminants, including conservation materials University of Oxford and the Department of Evolutionary Genetics at the Max Planck Institute for Evolutionary Anthropology, Leipzig, for their support. The (unless collagen-based glue has been applied). research leading to these results has received funding from the European Research Council (ERC) under the European Union’s Seventh Framework Pro- DNA Analysis. gramme (FP7/2007-2013); ERC Grant 324139 “PalaeoChron” (to T.H.) and ERC DNA extraction and library preparation. Genomic analysis of the four human Grant 694707 “100 Archaic Genomes” (to S.P.). We also acknowledge the bones was performed at the Max Planck Institute for Evolutionary Anthro- support of the Royal Society for funding a University Research Fellowship pology in Leipzig, Germany. Three samples (Vi-207, Vi-208, and Vi-*28) were (to M.B.).

1. Mellars P (2006) A new radiocarbon revolution and the dispersal of modern humans 12. Karavanic I, Smith FH (1998) The Middle/Upper Paleolithic interface and the re- in Eurasia. Nature 439:931–935. lationship of Neanderthals and early modern humans in the Hrvatsko Zagorje, Cro- 2. Hublin J-J (2015) The modern human colonization of western Eurasia: When and atia. J Hum Evol 34:223–248. where? Quat Sci Rev 118:194–210. 13. Smith FH, Trinkaus E, Pettitt PB, Karavanic I, Paunovic M (1999) Direct radiocarbon 3. Mellars P (2004) Neanderthals and the modern human colonization of Europe. Nature dates for Vindija G(1) and Velika Pecína hominid remains. Proc Natl – 432:461 465. Acad Sci USA 96:12281–12286. 4. Higham T, et al. (2014) The timing and spatiotemporal patterning of Neanderthal 14. Higham T, Ramsey CB, Karavanic I, Smith FH, Trinkaus E (2006) Revised direct radio- – disappearance. Nature 512:306 309. carbon dating of the Vindija G1 Upper Paleolithic Neandertals. Proc Natl Acad Sci USA 5. Fu Q, et al. (2015) An from with a recent Neanderthal 103:553–557. – ancestor. Nature 524:216 219. 15. Kozłowski JK (1996) Cultural context of the last Neanderthals and early modern hu- 6. Higham T (2011) European Middle and Upper Palaeolithic radiocarbon dates are of- mans in central-eastern Europe. The Lower and the Middle Palaeolithic, eds Bar- ten older than they look: Problems with previous dates and some remedies. Antiquity Yosef O, Cavalli-Sforza LL, March RJ, Piperno M (International Union for Prehistoric 85:235–249. and Protohistoric Sciences, Forli, ), Vol 5, pp 205–218. 7. Ovchinnikov IV, et al. (2000) Molecular analysis of Neanderthal DNA from the 16. Zilhao J, D’Errico F (1999) Reply, in the Neanderthal problem continued. Curr northern . Nature 404:490–493. – 8. Hublin J-J, Barroso Ruiz C, Medina Lara P, Fontugne M, Reyss J-L (1995) The Mous- Anthropol 40:355 364. 17. Zilhão J (2009) Szeletian, not Aurignacian: A review of the chronology and cultural terian site of Zafarraya (Andalucia, Spain): Dating and implications on the Palaeolithic peopling processes of western Europe. C R Acad Sci II 321:931–937. associations of the Vindija G1 Neandertals. Sourcebook of Paleolithic Transitions: 9. Pinhasi R, Higham TFG, Golovanova LV, Doronichev VB (2011) Revised age of late Methods, Theories, and Interpretations, eds Camps M, Chauhan P (Springer New – Neanderthal occupation and the end of the Middle Paleolithic in the northern Cau- York, New York), pp 407 426. casus. Proc Natl Acad Sci USA 108:8611–8616. 18. Svoboda J (2001) La question szélétienne. Les industries a outils bifaciaux du 10. Wood RE, et al. (2013) Radiocarbon dating casts doubt on the late chronology of the paléolithique moyen d’Europe occidentale, ed Cliquet D (Eraul, Liège, ), pp Middle to Upper Palaeolithic transition in southern Iberia. Proc Natl Acad Sci USA 110: 221–330.  2781–2786. 19. Karavanic I, Smith FH (2013) Alternative interpretations of the Middle/Upper Paleo- 11. Wolpoff MH, Smith FH, Malez M, Radovciˇ c J, Rukavina D (1981) Upper Pleistocene hu- lithic interface at Vindija Cave (northwestern Croatia) in the context of central Europe man remains from Vindija cave, Croatia, Yugoslavia. Am J Phys Anthropol 54:499–545. and the Adriatic. Archaeol Ethnol Anthropol Eurasia 41:11–20.

Devièse et al. PNAS Early Edition | 5of6 Downloaded by guest on September 27, 2021 20. Brock F, Higham T, Ramsey CB (2010) Pre-screening techniques for identification of 36. Gansauge M-T, et al. (2017) Single-stranded DNA library preparation from highly samples suitable for radiocarbon dating of poorly preserved bones. J Archaeol Sci 37: degraded DNA using T4 DNA ligase. Nucleic Acids Res 45:e79. 855–865. 37. Meyer M, Kircher M (2010) Illumina sequencing library preparation for highly mul- 21. Miracle PT, Mauch Lenardic J, Brajkovic D (2010) Last glacial climates, “Refugia”, and tiplexed target capture and sequencing. Cold Spring Harb Protoc, 10.1101/pdb. faunal change in southeastern Europe: Mammalian assemblages from Veternica, prot5448. Velika pecina, and Vindija (Croatia). Quat Int 212:137–148. 38. Dabney J, Meyer M (2012) Length and GC-biases during sequencing library amplifi- 22. Gansauge M-T, Meyer M (2014) Selective enrichment of damaged DNA molecules for cation: A comparison of various polymerase-buffer systems with ancient and modern – ancient genome sequencing. Genome Res 24:1543 1549. DNA sequencing libraries. Biotechniques 52:87–94. – 23. Krings M, et al. (2000) A view of Neandertal genetic diversity. Nat Genet 26:144 146. 39. Maricic T, Whitten M, Pääbo S (2010) Multiplexed DNA sequence capture of mito- 24. Green RE, et al. (2010) A draft sequence of the Neandertal genome. Science 328: chondrial genomes using PCR products. PLoS One 5:e14004. – 710 722. 40. Fu Q, et al. (2013) DNA analysis of an early modern human from , . – 25. Davies W, Hedges REM (2008 2009) Dating a type site: Fitting Szeleta Cave into its Proc Natl Acad Sci USA 110:2223–2227. – – regional chronometric context. Praehistoria 9 10:35 45. 41. Kircher M, Sawyer S, Meyer M (2012) Double indexing overcomes inaccuracies in 26. Markó A (2013) Leaf-shaped lithic and osseous tools from old excavated cave sites: multiplex sequencing on the Illumina platform. Nucleic Acids Res 40:e3. Demonstrating associations. The Sound of Bones, ed Lang F (Universität Salzburg, 42. Renaud G, Stenzel U, Kelso J (2014) leeHom: Adaptor trimming and merging for Il- Salzburg, ), pp 191–202. lumina sequencing reads. Nucleic Acids Res 42:e141. 27. Buckley M, Collins M, Thomas-Oates J, Wilson JC (2009) Species identification by 43. Li H, Durbin R (2009) Fast and accurate short read alignment with Burrows-Wheeler analysis of bone collagen using matrix-assisted laser desorption/ionisation time-of- transform. Bioinformatics 25:1754–1760. flight mass spectrometry. Rapid Commun Mass Spectrom 23:3843–3854. 44. Meyer M, et al. (2012) A high-coverage genome sequence from an archaic 28. Brown S, et al. (2016) Identification of a new hominin bone from , Si- individual. Science 338:222–226. beria using collagen fingerprinting and mitochondrial DNA analysis. Sci Rep 6:23559. 45. Green RE, et al. (2008) A complete Neandertal mitochondrial genome sequence de- 29. Richter KK, et al. (2011) Fish ’n chips: ZooMS peptide mass fingerprinting in a 96 well termined by high-throughput sequencing. Cell 134:416–426. plate format to identify fish bone fragments. J Archaeol Sci 38:1502–1510. 46. Welker F, et al. (2016) Palaeoproteomic evidence identifies archaic hominins associ- 30. Buckley M, Gu M, Shameer S, Patel S, Chamberlain AT (2016) High-throughput col- ated with the Châtelperronian at the . Proc Natl Acad Sci USA 113: lagen fingerprinting of intact microfaunal remains; a low-cost method for dis- – tinguishing between murine rodent bones. Rapid Commun Mass Spectrom 30: 11162 11167. 805–812. 47. Bronk Ramsey C (2001) Development of the radiocarbon calibration program. – 31. Buckley M, Harvey VL, Chamberlain AT (2017) Species identification and decay as- Radiocarbon 43:355 363. sessment of Late Pleistocene fragmentary vertebrate remains from Pin Hole cave 48. Reimer PJ, et al. (2013) IntCal13 and Marine13 radiocarbon age calibration curves – – (, UK) using collagen fingerprinting. Boreas 46:402–411. 0 50,000 years cal BP. Radiocarbon 55:1869 1887. 32. Marom A, McCullagh JSO, Higham TFG, Sinitsyn AA, Hedges REM (2012) Single amino 49. Stuiver M, Polach HA (1977) Discussion: Reporting of C-14 data. Radiocarbon 19: – acid radiocarbon dating of Upper Paleolithic modern humans. Proc Natl Acad Sci USA 355 363. 109:6878–6881. 50. Coplen TB (1994) Reporting of stable hydrogen, carbon, and oxygen isotopic abun- 33. Dabney J, et al. (2013) Complete mitochondrial genome sequence of a Middle dances (technical report). Pure Appl Chem 66:273–276. Pleistocene cave bear reconstructed from ultrashort DNA fragments. Proc Natl Acad 51. Brock F, Higham T, Ditchfield P, Bronk Ramsey C (2010) Current pretreatment Sci USA 110:15758–15763. methods for AMS radiocarbon sating at the Oxford Radiocarbon Accelerator Unit 34. Korlevic P, et al. (2015) Reducing microbial and human contamination in DNA ex- (ORAU). Radiocarbon 52:103–112. tractions from ancient bones and teeth. Biotechniques 59:87–93. 52. Marom A, McCullagh JSO, Higham TFG, Hedges REM (2013) Hydroxyproline dating: 35. Gansauge M-T, Meyer M (2013) Single-stranded DNA library preparation for the se- Experiments on the 14C analysis of contaminated and low-collagen bones. quencing of ancient or damaged DNA. Nat Protoc 8:737–748. Radiocarbon 55:698–708.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1709235114 Devièse et al. Downloaded by guest on September 27, 2021